The NCAR GPS dropsonde system, also known as AVAPS (Airborne Vertical Atmospheric Profiling System), is currently installed on about 21 aircrafts around the world. Each year about 5000 dropsondes are dropped over data-sparse regions such as over oceans or above remote mountain or polar regions to provide atmospheric thermo-dynamical and wind profiles. Since 1997, NOAA has used GPS dropsondes routinely during their hurricane reconnaissance flights to help predict the path and intensity of hurricanes. Lack of knowledge of performance of dropsonde humidity sensor has hampered the fully usage of the dropsonde humidity data. This study evaluates the performance of dropsonde humidity sensor using dropsonde data collected from three field experiments, DYCOMS-II (The Dynamics and Chemistry of Marine Stratocumulus Phase II: Entrainment Studies), IHOP (International H2O Project) and BAMEX (Bow Echo and Mesoscale Convective Vortex Experiment). The focus of the evaluation is on the dry bias of dropsonde humidity data suggested by previous studies and the performance of the dropsonde humidity sensor within clouds. The comparisons of dropsonde data with co-incident radiosonde data during IHOP show good agreements and suggest no dry bias in the dropsonde data. The IHOP dropsonde data will be compared with other IHOP humidity data. The behaviors of dropsonde relative humidity profiles were investigated using IHOP data when dropsondes were dropped above wave clouds. RH inside clouds sometimes reaches saturation, but sometimes does not, which is consistent with what we found with DYCOMS-II dropsonde data, in which all 63 dropsondes went through marine stratocumulus clouds. The DYCOMS-II dropsonde data are compared with humidity profiles obtained by air-borne dew-point hygrometer when the aircraft descended or ascended through the atmosphere. The new GPS dropsonde (Revision D, a new plastic clam shell nose piece, a hard sensor boom cover with desiccant to prevent the contamination dry bias, and a new one piece chute cap) was dropped during BAMEX in June and July, 2003. The BAMEX data will be analyzed to investigate the impacts of new sensor boom cover, which has been introduced to radiosonde since May 2000 but for the first time to dropsonde, and other new features for the new dropsonde.